Don't Feel Blue, Jet Blue

Inspiration

We were inspired to create this project out of our love for data science and learning. All four of us had little to no data experience coming in—something that we set about changing.

What it does

Our project is a dashboard that displays real-time insights with actionable feedback. The machine learning model is able to segment public opinion into more niche market categories, in order to determine the best course of action.

The model's current accuracy rate lies at about 85%.

How we built it

We first needed to determine if we were building a solution that'd address a real problem. We brainstormed potential single sources of failure and areas with unnecessary friction. Before proceeding further, we then asked real users. We created a survey and determined that there were definitely issues that users felt really impacted their airline experience. We consolidated these, and set out to build our product. We first developed low-fidelity sketches, before moving onto some tinkering, and finally our product.

We used pytorch and django for the backend. We used HTML/CSS/JS/Bootstrap for the frontend.

Challenges we ran into

Learning new things meant many challenges. Bhavesh had to learn REST. Michael and Jessica had to learn how to build the widgets in our dashboard. Jonathan had to learn how to handle the business side (jk, he already knows how to).

Accomplishments that we're proud of

The ML model is working!

What we learned

See above!

What's next for Don't Feel Blue, Jet Blue

Integration with a JetBlue API? It can potentially reduce friction if we made it part of a consumer-facing web app by allowing users to see what positive metrics the company meets. Social brand exposure is valued at about $8 per user and we believe that our solution can help direct these answers.